From language models to distributional semantics
Time: 01:00pm - 02:00pm
Place: Memorial Hall 401
Chung-chieh ShanDistributional semantics represents what an expression means as a vector that summarizes the contexts where it occurs. This approach has successfully extracted semantic relations such as similarity and entailment from large corpora. However, it remains unclear how to take advantage of syntactic structure, pragmatic context, and multiple information sources to overcome data sparsity. These issues also confront language models used for statistical parsing, machine translation, and text compression.Thus, we seek guidance by converting language models into distributional semantics. We propose to convert any probability distribution over expressions into a denotational semantics in which each phrase denotes a distribution over contexts. Exploratory data analysis led us to hypothesize that the more accurate the expression distribution is, the more accurate the distributional semantics tends to be. We tested this hypothesis on two expression distributions that can be estimated using a tiny corpus: a bag-of-words model, and a lexicalized probabilistic context-free grammar a la Collins.
In category: Computational linguistics
First and second-language patterns of variation: Acquisition and use of simple present and present progressive forms in Spanish and English
Time: 12:15pm - 01:00pm
Place: Ballantine Hall 004
Stephen FafulasThis investigation addresses the acquisition of the present progressive, and the variation of this form with the simple present, by English-speaking learners of Spanish in the US. Empirical research on the variation of these forms by native speakers (NS) of English and Spanish, and in particular the acquisition of the present progressive by nonnative speakers (NNS) of Spanish, is scarce. While seeking to fill this void in the literature, a second major contribution of this project is the extension of the analysis to the progressive constructions: seguir, venir, ir, and andar + V-ndo, which have received considerably less attention than the more frequent estar + V-ndo form, especially in the second-language (L2) variationist literature. Relatively little is known about how English-speaking learners of Spanish deal with the differences between the uses of the simple present and present progressive forms in these two languages. At what level of proficiency do they begin to employ both forms in similar contexts as NS of Spanish? Do they acquire the full range of progressive constructions, and sensitivity to the same linguistic predictors of use, found in NS grammars? The current study will empirically answer these questions through the use of film narrations, written contextualized questionnaires, and surveys of language-learning experiences from 115 participants. Results indicate that as learners of Spanish gain proficiency they move toward a second-language grammar which permits variation between the simple present and present progressive forms, as is found in native speaker grammars. However, while learners at the highest levels acquire the linguistic constraints guiding native speaker selection of the simple present and present progressive estar + V-ndo form, they do not acquire the full range of lexical forms used to express progressive aspect found in monolingual Spanish. These findings have implications for current contact-induced and interlanguage theories on Spanish-English grammatical development.
In category: Second language acquisition
Old Occitan as a verb-second language: The state of the argument
Time: 02:30pm - 03:30pm
Place: Ballantine Hall 242
Barbara VanceOne of the most far-reaching discoveries of late 19th-century Romance philology was the recognition that Old French (OFr), like Modern German, had a word-order constraint placing the finite verb in the second position of the clause. In the late 20th century, theoretical tools became available to explain this constraint in terms of verb-movement and, consequently, to account for the asymmetry between main clauses, which demonstrate the verb-second (V2) effect, and subordinate clauses, where the effect is largely blocked. It has traditionally been assumed that Old Occitan (OOc) is also a verb-second language because it shares many word-order patterns with Old French. However, demonstrating theoretically the V2 effect for Old Occitan is considerably more difficult than for Old French, for several reasons (e.g. OOc has a larger number of main-clause types that do not, strictly speaking, observe V2 order, and OOc lacks the atonic subject pronouns that mark the boundary between the basic (TP) and expanded (CP) areas of the clause in OFr). In this talk I present two aspects of OOc syntax that lend support to a verb-second analysis of the language, despite some recent attempts to dissociate OOc, and in some cases Old French as well, from the V2 type instantiated by German. We will look at (a) the complex alternation between tonic and null subject pronouns, which makes sense only as a reflex of a V2 system and (b) the postposition of object pronouns, a phenomenon that not only suggests the presence of an asymmetric verb-movement rule similar to that of (V to C in) Old French but also declines during the 13th and 14th centuries in tandem with the decline of verb movement in French. These characteristics point to the value of continuing research on verb-second syntax as an areal phenomenon in western Europe, while also encouraging attention to the wide range of variation found among the languages at issue (Germanic as well as Romance).
In category: Morphosyntax and semantics
JEvents v3.0.9 Stable
Copyright © 2006-2013
Powered by Joomla!®
Copyright © 2008 The Trustees of Indiana University | Copyright Complaints